Human biases and remedies in AI safety and alignment contexts

Zoé Roy-Stang; Jim Davies · 2025 · AI and Ethics 5:4891-4913 background low priority coded

Main argument

Thesis: cognitive biases (reviewed across public perception of AI, developer decision-making, and governance contexts) can undermine AI safety and alignment efforts; catalogues relevant biases and matching remedies - 'information consumer remedies' applicable at the individual level and 'information system remedies' incorporable into institutional design - to improve resource allocation, prioritization, and planning.

Why it matters here

Cognitive-bias catalogue for AI risk perception and safety decision-making, with individual and system-level debiasing remedies. Marginal to the dissertation's core, but a methodological resource for interpreting the folk corpus (which biases shape public AI discourse) and for the psychology of the stakeholder data.

Reading notes

Compact treatment (Carleton). Abstract read.

Roy-Stang, Z., & Davies, J. (2025). Human biases and remedies in AI safety and alignment contexts. AI and Ethics, 5, 4891-4913.

Close reading — 1 coded units

#1 · pp. 4891 · claim

“We discuss how relevant cognitive biases could affect the general public's perception of AI developments and risks associated with advanced AI. We focus on how biases could affect decision-making in key contexts of AI development, safety, and governance. We review remedies that could reduce or eliminate these biases.”

TU-METHOD ?

Synthesis-matrix row

Memos (1)

thesis-link · unit #1

Methodological accessory for the corpus chapters: when interpreting folk-comment distributions (e.g. availability-driven spikes after incidents, anchoring on sci-fi frames), Roy-Stang & Davies provide the bias taxonomy for a 'discourse hygiene' subsection - which folk patterns reflect stable normative positions vs known perception biases. Also supports the Brophy-style FILTRATION story: bias-aware filtering of folk judgments into considered judgments is precisely the CMJ step. One-cite use.